Therapy control using motion prediction based on cyclic motion model

1. An image-guided therapy delivery system, comprising:

a therapy generator configured to generate a therapy beam directed to a time-varying therapy locus, within a therapy recipient, the therapy generator comprising a therapy output to direct the therapy beam according to a therapy protocol;

an imaging input configured to receive imaging information about a time-varying target locus within the therapy recipient; and

a therapy controller configured to:

determine a latency associated with the therapy generator;

automatically generate a predicted target locus by applying an earlier target locus extracted from the imaging information and the determined latency associated with the therapy generator to a cyclic motion model associated with the therapy recipient; and

automatically generate an updated therapy protocol to align the time-varying therapy locus with the predicted target locus.

2. The image-guided therapy delivery system of claim 1 , wherein the therapy controller is configured to automatically generate the predicted target locus by:

extracting information indicative of a feature from the imaging information, the feature corresponding to the earlier target locus;

determining a phase of the cyclic motion model corresponding to a location of the feature;

estimating a change in the location of the feature using a later phase of the cyclic motion model corresponding to a scheduled time of an upcoming therapy delivery and the determined latency; and

applying information indicative of the estimated change in the location of the feature to provide the predicted target locus.

3. The image-guided therapy delivery system of claim 2 , wherein the therapy controller is configured to automatically generate the predicted target locus by:

identifying a boundary of the earlier target locus and determining a spatial point of the earlier target locus according to the identified boundary; and

relocating the determined spatial point of the earlier target locus using information indicative of the change in the location of the feature.

4. The image-guided therapy delivery system of claim 3 , wherein identifying the boundary of the earlier target locus includes using information about a difference in intensity between adjacent pixels or voxels.

5. The image-guided therapy delivery system of claim 1 , wherein the cyclic motion model specifies a motion of the target locus in at least one spatial dimension with respect to time or phase.

6. The image-guided therapy delivery system of claim 5 , wherein:

the cyclic motion model is established at least in part using a series of two or more acquisitions of three-dimensional imaging information of a region; and

the imaging input is configured to receive at least one-dimensional imaging information about the time-varying target locus within the therapy recipient after establishing the cyclic motion model.

7. The image-guided therapy delivery system of claim 1 , wherein the therapy controller is configured to control one or more actuators coupled to a moveable platform configured to support the therapy recipient, wherein the moveable platform is moved according to the time-varying locus.

8. The image-guided therapy delivery system of claim 1 , wherein the therapy controller is configured to control one or more apertures of a multi-leaf collimator, the apertures configured to shape the therapy beam based on the time-varying locus.

9. The image-guided therapy delivery system of claim 1 , wherein the therapy controller is configured to control one or more actuators coupled to the therapy output, the actuators configured to position the therapy output to establish a specified therapy beam direction.

10. The image-guided therapy delivery system of claim 1 , wherein the latency is a time period between an earlier image acquisition at a first time and a scheduled upcoming therapy delivery at a second time.

11. A method of adapting a therapy protocol in response to a time-varying target locus in an image-guided therapy delivery system, the method comprising:

receiving imaging information about the time-varying target locus within a therapy recipient;

determining a latency associated the a therapy generator of the therapy delivery system;

automatically generating a predicted target locus by applying an earlier target locus extracted from the imaging information and the determined latency associated with the therapy generator to a cyclic motion model associated with the therapy recipient; and

automatically generating an updated therapy protocol to align the therapy locus with the predicted target locus, the therapy locus established by a therapy beam provided by the therapy generator.

12. The method of claim 11 , wherein automatically generating the predicted target locus comprises:

extracting information indicative of a feature from the imaging information, the feature corresponding to the earlier target locus;

determining a phase of the cyclic motion model corresponding to location of the feature;

estimating a change in the location of the feature using a later phase of the cyclic motion model corresponding to a scheduled time of upcoming therapy delivery and the determined latency; and

applying information indicative of the estimated change in the location of the feature to provide the predicted target locus.

13. The method of claim 12 , wherein automatically generating the predicted target locus comprises:

identifying a boundary of the earlier target locus and determining a spatial point of the earlier target locus according to the identified boundary; and

relocating the determined spatial point of the earlier target locus using information indicative of the change in the location of the feature.

14. The method of claim 13 , wherein identifying the boundary of the earlier target locus includes using information about a difference in intensity between adjacent pixels or voxels.

15. The method of claim 12 , wherein the cyclic motion model specifies a motion of the target locus in at least one spatial dimension with respect to time or phase.

16. The method of claim 12 , comprising:

establishing the cyclic motion model at least in part using a series of two or more acquisitions of three-dimensional imaging information of a region; and

after establishing the cyclic motion model, receiving at least one-dimensional imaging information about the time-varying target locus.

17. The method of claim 11 , wherein the latency is a time period between an earlier image acquisition at a first time and a scheduled upcoming therapy delivery at a second time.

18. A medical system, comprising:

a medical device including a control protocol, the control protocol being configured to control the medical device; and

a controller configured to:

receive a measured data set containing values of a physiologic signal caused by an anatomical structure on a patient measured at a plurality of past time points;

estimate a state representation for each of the past time points according to the measured data set, the state representation reflecting an internal state of the anatomical structure;

determine a latency associated with the medical device;

predict a future state representation for a future time point based on the estimated state representations and the determined latency associated with the medical device;

train a mapping function using historical measured values of the physiologic signal;

predict a functional representation for the future time point based on the future state representation and the mapping function mapping state representations to functional representations of the physiologic signal;

calculate a future value of the physiologic signal according to the predicted functional representation, and

update the control protocol according to the future value.

19. The medical system of claim 18 , wherein the internal state of the anatomical structure comprises a source of quasiperiodic motion.

20. A method for updating a control protocol of a medical device, comprising:

receiving a measured data set containing values of a physiologic signal caused by an anatomical structure on a patient measured at a plurality of past time points;

estimating a state representation for each of the past time points according to the measured data set, the state representation reflecting an internal state of the anatomical structure;

determining a latency associated with the medical device;

predicting a future state representation for a future time point based on the estimated state representations and the determined latency associated with the medical device;

training a mapping function using historical measured values of the physiologic signal;

predicting a functional representation for the future time point based on the future state representation and the mapping function that maps state representations to functional representations;

calculating a future value of the physiologic signal according to the predicted functional representation; and

updating the control protocol of the medical device according to the predicted future value, the control protocol controlling the medical device.